If the only goal is producing electricity at the current minimum price per kWh, then you have a good point.

OTOH if you also have a long-term goal of figuring out how to effectively design and build fusion reactors, then it's worthwhile to build them as best you can even if wind is currently more cost-effective.

As for why you might have such a long-term goal, I can think of several reasons:

1. Future fusion reactor designs might be much more economical to build and run, once enough hands-on experience has been gathered to make them so
2. Fusion reactors could be used in situations where wind power isn't available (e.g. space exploration)

Certainly the Secret Service would love to have an automatic head's-up on the true identity of anyone walking into the room. Not that I would expect this implementation to be reliable enough actually to be used that way, but someday one will be.

If I was deliberately trying to cause a nation-wide backlash against data centers, I'm not sure what I'd be doing differently from what the AI companies are currently doing.

Has nobody told them that people don't like having their lives disrupted, particularly when they don't see any compensating benefit, or even a convincing reason for having any of it? If they were to ease off the gas pedal just a bit, they could probably do a boil-the-frog and get a larger number of smaller/less-obtrusive data centers built over a longer time period, and without the voter revolts and strict legislation that are likely to hobble them now.

Objective C's syntax is objectively terrible; it mixes Smalltalk syntax with C syntax, confusing everyone. Its implicit heap usage makes it unusable for real-time programming (e.g. even Apple had to fall back to C to implement CoreAudio), and ARC is only a partial substitute for RAII.

Apple was right to switch over to Swift.

That's true, but when it takes Joe Random Hacker 10 seconds to generate a plausible-looking pull-request, which requires Joe Project Maintainer to spend 30 minutes reviewing the code-changes in that request, and Joe Project Maintainer isn't getting paid for his time spent doing the review, you've got all the ingredients for a distributed-denial-of-service attack on your project's maintainers. Perhaps AI code-reviewers can restore the balance, but I don't know how many project maintainers would trust their codebase's integrity to them (yet).

If they left them in place they might be used to collect data, which might in turn be used to show that climate change is a real problem. We can't have that, it gets in the way of maximizing oil companies's profits.

It's certainly possible to translate COBOL source into another computer language of your choice, although I'm not sure LLMs are the best tool for that job. An LLM might be able to give you more readable post-translation source code, but traditional machine translation would give you post-translation source code that works correctly, which is probably more important.

Another way to phrase that would be, "This is why C code is bad, because C programmers are expected to understand the rules about how to not leak memory, but there is no mechanism to enforce that requirement".

... and to their credit, eventually some of them do figure it out, and after that they (mostly) write good C code that doesn't leak. However, that doesn't change the fact that at any given moment there are millions of unseasoned C programmers out there who haven't reached that point yet, and who are nevertheless writing leaky code which gets put into production and causes trouble; and new C programmers appear every day. It's the Eternal September problem, applied to memory management.

So either (a) we ban C programmers from pushing to production until they've had at least 5 years of experience, or (b) we find some means to flag their errors at build-time, or (c) we live with the status quo messiness indefinitely. Linux is going with Rust as their mechanism for implementing plan (b).

If you have access to a God-tier LLM that you can rely on to find every bug, I think that could work.

However, I don't think anyone in the Linux community is ready to trust LLMs to that extent just yet. Not only are they quite fallible, they are also non-deterministic -- so if you ask your favorite LLM to find the bugs in the code, and it doesn't find any, and then you feed it the exact same prompt again, it might find some on its second attempt. So how do you know when to stop re-asking?

LLMs are currently constituted are very useful for finding bugs, but not so useful for guaranteeing that no bugs remain.

You know what improves code quality? Process improvements.

You know what doesn't improve code quality? Telling people that they suck. It's pointless and immature.

They might get a better responsie by requesting $4.99/month to remove the Meta AI slop from their Facebook interface.

Perhaps, but I suspect Starlink (etc) already fills most of that use-case, and for the rest, they'll want that compute to be physically located inside the drones themselves, because otherwise the drones will be susceptible to jamming or spoofing.

There will be jobs that AI can't do. How many? Enough to keep 5-10 billion humans employed? What makes you so sure there will be?

Clearly AI has progressed considerably over the last 5-10 years. It's anyone's guess how much further it will progress -- maybe it'll plateau right where it is now, or maybe it will keep becoming more powerful as better algorithms are discovered. I'm not qualified to predict that, and neither are you, but the AI people certainly seem bullish about it.

Sorry, I thought you would understand that I meant that the resources that money represents appear, once you've solved the automation problems that currently make mass-production difficult. That's why you can buy a pocket computer today for $300 that would have your cost you billions of dollars twenty years ago, if you could have obtained it at all.

I bet you'd have more interesting conversations if you made a good-faith effort to understand what the other person was saying, rather than just jumping straight to the part where you get to throw insults at them and tell them how dumb they are. Doesn't that get boring?

I have yet to hear of a remotely plausible business case for putting data centers into space. The only benefit is 24/7 solar power, but that benefit is more than offset by the cost of launching everything into orbit, plus the cost of keeping everything properly cooled, plus the cost of radiation-hardening everything, and finally the cost of maintaining hardware in space (or, more likely, the cost of periodically having to write off the entire investment and build and launch new replacement hardware).

Unless Musk is trying to corner the market for AI-generated kiddie-porn (or something similarly illegal that needs to be operated beyond the reach of Earthly authorities), his ground-based competitors will undercut his pricing by a factor of 100, and he therefore won't have a viable product to sell.

Certainly new categories of work will emerge. The question is, will hiring and paying human beings be the most economically efficient way to fill those new positions, or will those jobs be done by AIs instead?

Previous waves of automation allowed people to move "up the food chain" and do jobs the machines still couldn't do, which was fine (at least, for the people capable of doing the new jobs), but if we get to the stage where there aren't many jobs left that the machines can't do, then we're out of luck -- it's unlikely that our tech-bro overlords are going to hire people simply on humanitarian grounds, if they can get an unquestioning machine to do the same work cheaper.

I agree that UBI is unlikely, but only because the billionaires don't like sharing and therefore won't support it. The money does "just appear" when you have mass automation doing the work to make it appear, but it will go into Bezos' checking account, not to the general public.